yusufipek
80ff4ecade
Three files disagreed. The test workflow gained a job on either side, so both stay: the Mac now parses the release and packaging scripts as well, and Windows keeps its own job below. `restart` and `launch_gui` both start Dikte again, and master moved that argv into `ipc.launcher()` because a packaged build has no `__main__.py` to name. Windows still cannot use execv there, so the detached start it needs now takes what the launcher hands it rather than spelling the interpreter and the script itself. The test counts in CONTRIBUTING are the suite as it stands after the merge: 1104 of 1147 run anywhere, and the 43 left are the Linux ones.
5.0 KiB
Contributing
Running the tests
python -m unittest discover # all of them, about a second
python -m unittest tests.test_api # one file
python -m unittest tests.test_api.Transcribe.test_no_key_at_all
Nothing to install: the tests use the standard library's unittest, and the
only dependency is the PyQt6 the application already needs. They reach neither
the network, the microphone, nor your real ~/.config/dikte, so they are safe
to run anywhere and they run on a machine with no display.
CI runs the same command on Python 3.11 through 3.13. A pull request that turns it red will not be merged.
Writing one
Put it in tests/, named after the module it covers. Inherit from
tests.support.DikteTest whenever the code under test touches a file, a
setting or the interface language: it hands the test its own config and data
directories, resets the language, and puts them back afterwards.
tests/support.py has the rest of what you need:
| For | Use |
|---|---|
| An HTTP call | fake_urlopen(reply, …), then read the recorded requests |
| A reply that fails | http_error(429), url_error(), raw_body("not json") |
| Reading what was sent | sent_json(request), multipart_fields(request) |
| A program on the PATH | only_these_tools("pactl", "wl-copy") |
| Standing on another system | mock.patch.object(sys, "platform", "darwin") |
| Audio | silence(), tone(), speech(), stereo(), make_wav() |
| A settings object | self.config(cleanup_enabled=False) |
Three things about this codebase trip up a new test:
Signals from a worker thread are never delivered. Pipeline, MeetingPipeline
and FileTranscriber emit from the thread start() spawned, which Qt queues
until an event loop runs one. Call _work() directly instead: it is the same
code one frame down, and the signals arrive at once.
A level that never moves is not speech. The silence check is relative, so a
steady tone reads as its own noise floor however loud it is. Use speech()
rather than tone() when a recording is meant to have somebody talking in it.
cli.launch_gui replaces the process. With no instance running, some verbs
os.execv into the application, which would take the test run with it. Patch
cli.launch_gui. DikteTest blocks os.execv as a backstop, so a test that
forgets fails rather than hangs.
Another platform
Four systems are supported: Wayland, X11, macOS and Windows. Each one is a named
entry in a table, and one chooser picks between them, so a fifth adds an entry
and a line rather than a branch inside every function. The tables are
paste.Desktop (clipboard and key press), audio.Sound (capture, the device
lists, and whether the far side of a meeting can be recorded at all),
paths.directories() (where the settings and the data live) and hotkey.backend(),
which names the one mechanism a session has for holding a key. Keep sys.platform
inside the chooser and read it there every time: a constant settled at import is
one no test can stand somewhere else.
Where a platform cannot do something, say so in its table entry rather than in
the code that asks. audio.Sound.meetings is the shape of it: Windows offers no
capture device for what the speakers are playing, and a caller reading a False
there can tell that apart from an empty device list, which only means the tool
that lists them is not installed.
The tests are split along the same line, and almost none of them are skipped.
1104 of the 1147 run on any machine, including every line of the Wayland, X11,
macOS and Windows backends: the programs are faked at shutil.which, the
frameworks and system libraries at the one function that loads them
(paste._win_api, hotkey._win_input). A test class says which system it is
standing on rather than avoiding the question:
class MacOS(ClipboardContract, DikteTest):
platform = "darwin"
here = paste.MACOS
so the Linux half is checked on a Mac, the macOS half on Linux and the Windows half on both, and a change to a chooser cannot quietly break the platform nobody is sitting at. What the systems owe in common is written once as a contract class and subclassed by each of them.
The 43 that do carry @linux_only are the ones that would need the real thing:
the /dev/input listener, KDE's shortcut file, GNOME's gsettings. Mark a test
that way only when faking it would leave nothing to test. A test that quietly
stops running on the platform you are porting to protects nothing.
What a pull request should carry
A change to behaviour comes with a test for it. Adding a provider means a row in
config.TRANSCRIBERS and a test that the request goes to the right URL with the
right fields; adding a platform means a test for whatever the parsing of its
device list, clipboard or shortcuts looks like. Adding a setting means both halves of settings_ui.py: the round
trip in tests/test_ui.py is what catches only one of them being written.
Match the surrounding code: it is plain Python with no framework, comments explain why rather than what, and neither the code nor the commit messages use an em dash.